proposal of a permanent magnet hybrid-type axial

IEEJ Journal of Industry ApplicationsVol.4 No.4 pp.339–345 DOI: 10.1541/ieejjia.4.339

Paper

Proposal of a Permanent Magnet Hybrid-typeAxial Magnetically Levitated Motor

Nobuyuki Kurita∗a)Member, Takeo Ishikawa∗ Member

Hiromu Takada∗ Non-member, Genri Suzuki∗ Non-member

(Manuscript received July 30, 2014, revised March 11, 2015)

A permanent magnet hybrid-type axial magnetically levitated motor is proposed in this paper. The motor consistsof a spherical permanent magnet and an axial type bearingless motor with a tilt control function. The rotor has twopermanent magnets on one side, and the motor stator has eight poles, which include eight concentrated windings. Theoperating principle was investigated through numerical analysis. It was verified that the proposed motor can control thetranslational motion, inclinational motion, and rotational motion independently. A test rig was fabricated to investigatethe control performance. The time responses of each controlled axes were quick. In addition, stable levitated rotationwas observed in each actively controlled axes up to 2000 min−1.

Keywords: axial flux motor, bearingless motor, magnetically levitation technology, passive magnetic bearing

1. Introduction

A magnetically levitated motor (maglev motor) can sup-port and rotate a rotor by using magnetic force with no me-chanical contact and, therefore, has many advantages overconventional mechanical bearings (1) (2). In order to accom-plish complete magnetic levitation of the rotor, however, it isnecessary to control five axes actively. Two radial magneticbearings and one axial magnetic bearing are usually required,which gives rise to major issues, such as a complicated levi-tation control system and associated electronics that result inenlarged equipment size.

In order to down size a maglev motor and to simplify thecontrol system, several types of bearing-less motors (B�M)have been proposed and researched, such as a radial-B�M (3) (4)

and an axial-B�M (5) (6). To increase passive stability, the rotorshape should be a disk shape for radial-B�M or a long cylin-drical shape for axial-B�M. Consequently, permanent mag-net (PM) size and stator surface area will be decreased due tothese rotor shapes. Additionally, these geometries are associ-ated with low rotational torque and weak suspension force.

To solve this problem and increase PM size and stator sur-face area, several types of axial maglev motors that have in-clination motion control function were researched. Shimboet al. (7) and Chongk-wanyuen el al. (8) used an axial-B�M thathas a PM synchronous motor on the upper side and a reluc-tance motor on the lower side. Osa et al. (9) and Takada etal. (10) used an identical stator on both sides of the rotor. Sincethese axial-B�Ms required two stators to control the tilt mo-tion, the number of amplifiers required for levitation controlincreases, and the levitation control becomes more complex.

a) Correspondence to: Nobuyuki Kurita. E-mail: [email protected]∗ Gunma University

1-5-1, Tenjin-cho, Kiryu, Gunma 376-8515, Japan

Fig. 1. Schematic of proposed maglev motor

As an alternative, the research proposed in this paper de-tails a PM hybrid type axial maglev motor, shown schemat-ically in Fig. 1. The maglev motor consists of a sphericalpermanent magnet as a upper side stator and an axial-B�Mas a lower side stator, and a rotor. The spherical PM is fixedon housing in order to produce upward attractive force on therotor. The axial-B�M controls rotation, axial translation andtilt motion actively. A spherical permanent magnet was usedin order to produce axial directional upward attractive force(z) to levitate a rotor with zero-bias current. It is also usedto minimize the negative torque produced to tilt direction (θx,θy). A rotor disk diameter is bigger than the spherical perma-nent magnet diameter, so the attractive force of the permanentmagnet won’t limit movement of the rotor in radial direction(x, y). Moreover, magnetic pole of the spherical permanentmagnet doesn’t change when the rotor is rotating, so perma-nent magnet won’t produce negative rotating torque in rota-tional direction (θz). The number of amplifiers required isonly eight. This is half as many as compared with the doublestator type magnetically levitated axial gap motor (10). More-over, the structure of the proposed motor is very simple. The

c© 2015 The Institute of Electrical Engineers of Japan. 339

PM hybrid-type Maglev Motor（Nobuyuki Kurita et al.）

motor is proposed to develop a left ventricular assist deviceor a small fan in particular.

The operating principle is described below, as is an FEMmagnetic analysis that indicates the proposed maglev motorcan control each degree of freedom independently. Based onthis theoretical framework, an experimental setup was fabri-cated. Magnetic levitation was achieved and the control per-formance of the maglev motor was also investigated.

2. Operating Principle

The proposed maglev motor consists of a spherical PM anda conventional axial-B�M. The rotor of the axial-B�M hastwo crescent shape PMs, and the stator has eight salient poleswith a concentrated winding on each pole. The spherical PMas the upper side stator always produces upward attractiveforce, while the motor PMs always produce downward at-tractive force. Rotation, axial translation, tilt motion are con-trolled actively by the interaction between the magnetic fieldproduced by the motor PM and the magnetic field producedby the control currents. Figure 2 is top view of the maglevmotor. The spherical PM, the rotor disk, the rotor core, theback yoke and windings were omitted. Motor PM was de-picted translucently. Characters NPM and S PM written on themotor PMs means the magnetic pole at the lower side airgap.The flux density distribution produced by the motor PMs inlower-side airgap is described by Eq. (1).

Bpm(θ, t) = BPM cos(ωt + θ) · · · · · · · · · · · · · · · · · · · · · · (1)

where BPM(T) is the maximum value of the rotor PM fluxdensity, ω (rad/s) is rotation speed, t (sec) is time, and θ (rad)is angle. To control the rotor rotation θz, a two-pole rotatingmagnetic field of phase difference ψ (rad) is generated by thestator windings.

The flux density Bθz of the rotational control current can bewritten as (2)

Bθz(θ, t) = −BΘZ cos(ωt + θ + ψ) · · · · · · · · · · · · · · · · · (2)

where BΘZ is the maximum value of the flux density producedby the motor current.

A schematic of magnetic pole arrangements is shown inFig. 3(b). A two-pole rotating magnetic field of phase dif-ference ψ (rad) results in a rotating torque in a counter-clockwise direction θz.

To control the axial translational motion z, the stator wind-ings should produce a two-pole magnetic field which has thesame phase as the PM magnetic field. The flux density Bz

created by the axial direction control current would thus de-scribed as (3)

Bz(θ, t) = −BZ cos(ωt + θ) · · · · · · · · · · · · · · · · · · · · · · · (3)

where, BZ is the maximum value of the flux density producedby the axial direction control current. Figure 3(b) shows themagnetic pole arrangements of the rotor PM and the axialdirection control current. Here, the left side poles are mag-netized to N poles at the same time the right side poles aremagnetized to S poles, so an attractive force acts between therotor and the stator, resulting in a downward suspension forceFZ acting on the rotor. Meanwhile, to control the tilt motionθx and θy, a four-pole magnetic field should be generated by

Fig. 2. Top view of the maglev motor and coordinatesystem

(a) Rotation control (b) Translation control

(c) θx tilt motion control (d) θy tilt motion control

(e) Restoring force Fx (f) Restoring force Fy

Fig. 3. Magnetic pole arrangement

the control current. The flux density Bθx and Bθy for θx andθy tilt motion control are given by (4)

Bθx(θ, t) = BΘ sin(ωt + 2θ)Bθy(θ, t) = −BΘ cos(ωt + 2θ)

· · · · · · · · · · · · · · · · · · · · · (4)

where BΘ is the maximum value of the flux density producedby the tilt control current. In Fig. 3(c), the left near side poleand the right far side pole are magnetized to N poles, and theleft far side pole and the right near side pole are magnetizedto S poles. In this configuration, an attractive force acts on therotor at the near side and a repulsive force acts in far side. Asa consequence of these forces, a restoring torque τθx acts in θx

direction. A restoring torque τθy for θy direction is producedin the same way. In Fig. 3(d), magnetic pole arrangementfor τθy is depicted. Additionally, radial direction translationalmotions are stabilized passively. Radial directional restoringforce Frx, Fry for displacement x, y are shown in Figs. 3(e),

340 IEEJ Journal IA, Vol.4, No.4, 2015


Table 1. Principle sizes of the FEM model

(f). When rotor was displaced in x or the y-direction, a radiuscomponent of the attractive force between the rotor PM andstator core acts on rotor. This radial directional force is usedas restoring force.

3. Magnetic Field Analysis

3.1 Validation of Control Force Independency Inorder to verify that rotation control, translation control, andinclination control are not cross-coupled, an FEM magneticfield analysis was carried out; principle sizes of the modelare listed in Table 1. The rotor was located at the geometriccenter between the spherical PM and motor stator. The air-gap between the rotor and the stator was 3 mm. By applyingeach control current (rotation control current Iθz, axial direc-tion translational motion control current Iz, and tilt controlcurrents Iθx, Iθy) to each coil according to (2), (3) and (4),the rotation torque τθz, translation suspension force FZ andrestoring torques τθx, τθy were calculated.

When the phase difference of the stator pole and rotor poleis ψ = π/2, the relationships between rotational torque τθz andeach control current are as shown in Fig. 4(a). Specifically, alinear relationship exists between the rotational torque τθz androtation control current Iθz, but the rotational torque createdby translation control current and tilt control current are neg-ligibly small. Figure 4(b) shows the relationships betweenthe axial direction suspension force FZ and each control cur-rent. When the upper and lower airgaps are both 3 mm, theresulting attractive force by the motor PMs is stronger thanthe attractive force of the spherical PM by 1.8 N. The sus-pension force FZ is approximately proportional to the trans-lational motion control current Iz. With a low control currentof ±1.5 A, the suspension force produced by the rotation andboth tilt control currents are negligible. However, when thecurrent of more than ±1.5 A is applied, the rotation and θx tiltcontrol currents produce a suspension force. This force is dueto the interaction of the flux produced by the spherical PM onthe flux produced by the rotor PM, which results in the flux

(a) Rotation torque (b) Suspension force

(c) θx restoring torque (d) θy restoring torque

(e) x-directional force (f) y-directional force

Fig. 4. Analytical result of interference of control cur-rent

density distribution in the lower side airgap varying from theideal sinusoidal wave.

The relationship between restoring torque τθx and eachcontrol current are shown in Fig. 4(c). Restoring torque θx ismainly produced by tilt control current Iθx, however, the rota-tional motion control current Iθz does contribute some torque.Figure 4(d) shows a similar relationship between restoringtorque τθy and each control current. The variation of the fluxdensity distribution of the motor PM caused by the flux of thespherical PM produced a constant torque of 37 mNm in theθy direction. As seen for the other tilt direction, the restoringtorque θy is mainly produced by tilt control current Iθy, butsome contribution also comes from the translation motioncontrol current Iz. Those interferences of torque and forceare again due to the variation of the flux density distributionof the motor PM. We postulate that a closed loop magneticcircuit for the upper side PM would decrease the mutual in-terference of torque and force.

Finally, Figs. 4(e) and (f) show the relationship betweenradial directional force Fx, Fy and each control current. In-terferences of torque and force are observed. In this case x-directional force produced by translation motion control cur-rent Iz and y-directional force produced by the rotation con-trol current Iθz are thought to be due to the variation of theflux density distribution of the motor PM. Meanwhile, the x-directional force Fx produced by tilt control current Iθy andand y-directional force Fy produced by tilt control current Iθz



Fig. 5. Axial directional force variation with, displace-ment and control current

are thought to be due to the structure of maglev motor.Forces Fx, Fy in the radial direction were not observed in

the double stator type maglev motor (10) because they werecanceled due to the symmetric stator design. In a physicalsetup, perfect symmetry is not achievable, but the magnitudeof this force issmall enough such that it does not pose a signif-icant problem for magnetic levitation control; an additionalelectromagnet could be used to reduce the radial forces if de-sired.3.2 Validation of Controllable Range In order to

verify the controllable range of the rotor movement, anotherFEM magnetic field analysis was carried out. Initially, theairgap between the rotor and the stator was 3 mm and therotor was located at the geometric center between the spher-ical PM and motor stator. A suspension force coefficient of2.59 N/A was obtained at the geometric center. The rotor wasthen moved axially (z-direction) in the range of ±1 mm, andthe control current of ±3 A was applied in each position; ana-lytical results are shown in Fig. 5. According to the analyticalresult, it can be estimated that the negative 3 A control currentcan produce negative directional force even when the rotor isdisplaced toward the positive direction about 0.74 mm, andthe positive 3 A control current can produce positive direc-tional force when the rotor is displaced toward the negativedirection about −0.78 mm. Thus, the controllable rage of theaxial direction can be estimated from 0.74 mm to −0.78 mm.It should be noted that the controllable range in the positiveand negative direction was not same due to the difference be-tween the magnetic equivalent point and the geometric center.

Next, the rotor was tilted around x and y axes in the rangeof ±4.5 deg, and the control current of ±3 A was applied ineach angle; results are shown in Fig. 6. According to the an-alytical result of the restoring torque θx shown in Fig. 6(a),it can be estimated that ±2.23 A control current can producesufficient control torque when the rotor tilted about 4.5 deg.Thus, the controllable rage of the tilt motion can be estimatedfrom ±4.5 deg. A restoring torque coefficient of 32.2 mNm/Awas obtained at the geometric center. In contrast, the restor-ing torque coefficient for θy was 26.9 mNm/A at the geomet-ric center. As shown in Fig. 6(b), it can be estimated that anegative 2.74 A control current can produce negative direc-tional torque when the rotor tilts toward the positive directionabout 4.5 deg, and positive 3 A control current can producepositive directional torque when the rotor tilts toward the neg-ative direction about −1.79 deg. Thus, the controllable rangeof the restoring torque of θy can be estimated from 4.5 deg to−1.79 deg. Again, the asymmetry is due to the variation of

(a) Restoring torque around x axis

(b) Restoring torque around y axis

Fig. 6. Restoring torque variation with, tilt and controlcurrent

the flux density distribution of the motor PM caused by theflux of the spherical PM.

4. Experimental Results

4.1 Experimental Setup According to the analyticalresult, a simple experimental setup was designed and fabri-cated. Figure 7(a) shows the upper and lower stator. Theupper stator consists of spherical PM and PM holder, whilethe lower stator consists of the eight pole stator and concen-trated winding. Figure 7(b) shows top and bottom views ofthe rotor. The rotor disk is made of ferromagnetic materialin order to produce upward attractive force toward the upperstator; the disk surrounded by a sensor target installed aroundthe rotor core. Two crescent shaped PMs were added to theunderside produce magnetic fields for rotation and downwardattractive force. A Fig. 7(c) shows the entire unit of the fab-ricated test rig. Although omitted in the figure, three eddycurrent type gap sensors were evenly spaced along the ax-ial direction in order to detect axial directional displacementand tilt motion of the rotor. A control system of the mo-tor is shown in Fig. 8. The rotor displacement z1, z2 and z3

detected by three eddy current type gap sensors are used tocalculate axial displacement of the center of the gravity z andthe rotor tilts θx, θy. Calculated displacement signals are fedto a digital signal processor via AD converter. Three inde-pendent digital PID controllers for these displacement sig-nals are constructed by the DSP. The PID controllers cal-culate control current reference values for each axis. ThePID gains were tuned by trial and error method based onthe time responses of the rotor. The sampling time inter-val was t = 0.1 msec. Translation control gains for levitationcontrol were as follows; proportional gain KZP = 6.2 A/mm,derivative gain KZD = 0.015 A·sec/mm, Integrated gain KZI =

3.0 A/(mm·sec). Tilt control gains are follows; proportionalgain KθP = 5.0 A/deg, derivative gain KθD = 0.011 A·sec/deg,



(a) Photo of the upper and lower stator

(b) Photo of the rotor

(c) Entire unit

Fig. 7. Photo of the fabricated experimental setup

Fig. 8. Diagram of the control system of the motor

Integrated gain KθI = 3.0 A/(deg·sec). Although omitted inthe figure, forty nine small PMs were attached underneaththe sensor target. Forty eight PMs are attached evenly spacedand one PM is attached inner side of these PMs in order toworks as index. And three hall ICs (i.e. A-Phase, B-Phaseand Index) were installed to compose an angular encoder.4.2 Static Force Characteristics Static force in the

axial direction was measured and compared with the analyt-ical results. Figure 9 shows measurement results when therotor was displaced ±1.0 mm and a control current of ±2.4 Awas applied. The measured force was weaker than the ana-lytical result by about 20% to 30%. This is considered dueto the error of material characteristics and roughness of themesh size at the wide air gap of 3 mm. Suspension force co-

Fig. 9. Measured result of axial directional force, dis-placement and control current

efficient of 1.92 N/A was obtained at the geometric center.This value is 74.1% of the analytical result.

In accordance with the analytical result, the negative 2.4 Acontrol current could produce negative directional force evenwhen the rotor was displaced toward the positive directionabout 0.74 mm. However, the controllable range to the nega-tive directional movement became especially narrow by com-parison with the analytical result. The axial range was esti-mated to be 0.77 mm to −0.44 mm. The restoring torque wasnot measured but we expect it may be weaker than that pre-dicted by the analytical result, similar to the force results inthe axial direction.4.3 Levitation Control Performance Rotor levita-

tion control with 0 min-1 rotation was achieved. To controlthe axial translation motion, the tilt, and the rotation inde-pendently, eight sets of linear amplifiers were required. Oncethe rotor was levitated stably, an impulse disturbance was ap-plied, and the rotor displacement and control current wererecorded. Duration of an impulse disturbance signal wasfixed on 4 msec and the strength was changed. The rotorwas displaced 0.1 mm in axial direction, 0.5 deg in tilt θx di-rection, 1 mm in radial direction. As shown in Fig. 10(a),the response of the axial controller is fast; the rotor vibra-tion dropped to less than ±5% of the maximum displacementwithin about 0.02 sec in response to an impulse input whichmoved the rotor about 0.1 mm in the z-direction. Moreover,since there was almost no rotor vibration in the inclinationdirection (θx, θy) by the z-direction perturbation, the axial di-rection translation control does not appear to affect the con-trol performance of the tilt control. The response of the tiltcontroller was measured to be similarly quick for an impulsetilting input of about 0.4 deg in the θx-direction (Fig. 10(b));the rotor’s rotational vibration became less than ±5% of themaximum tilt within about 0.03 sec. It was also verified thatthe tilt control also does not affect the control performanceof the axial translation control. Figure 10(c) shows the rotorbehavior in response to an impulse input which moved therotor about 1.0 mm in the radial direction. In this case, theradial vibration took nearly 0.9 sec to fall below ±5% of themaximum displacement. The slow settling time of the radialdirection control is a consequence of the radial direction be-ing stabilized passively by the restoring force produced bythe axial attractive force (rather than employing an active sta-bilization control).4.4 Levitated Rotation Performance In order to

clarify levitated rotation characteristics, vibration amplitudein the axial translation, tilt and radial axes were measured



(a) Disturbance input: Axial direction z

(b) Disturbance input: Tilt θx

(c) Disturbance input: Radial direction x

Fig. 10. Impulse response result

at rotational speeds from 0 min−1 to 2000 min−1. The rotorspeed was increased stepwise by 100 min−1 after the rotorreached a steady state speed, and the vibration amplitude ofthe rotor was recorded (Fig. 11).

The rotor vibration of radial direction was increased atspeeds of 600 min−1 to 900 min−1 due to the weak restoringforce of the passive control in the radial direction. Although

(a) Axial direction z

(b) Radial direction: x

(c) Tilt: θx , θy

Fig. 11. Vibration amplitude

the increase in the vibration amplitude of the radial directionalso affected the vibration amplitude of the axial directionslightly, the active control in the axial direction is capable ofmaintaining stable, levitated rotation. In the rotating speedrange beyond 1300 min−1, however, levitation control of theaxial direction became unstable. Thereby, the vibration am-plitude of the radial direction also increased such that levi-tated rotation was broken at the speed of 2100 min−1. This isbecause the vibration amplitude of the axial direction becamelarger than the controllable range. The levitation control ofthe tilt directions was stable in all speed regions.

5. Conclusion

A permanent magnet hybrid type axial magnetically lev-itated motor was proposed. The FEM magnetic analysisshowed that there was no significant interference between ro-tational torque, suspension force, and restoring torque, indi-cating that these four axes are therefore able to be controlledindependently. The impulse response measured in the test rigshowed quick response for the actively controlled degrees offreedom (z and θx, θy) and showed that the translation motionand the inclination motion can be controlled independently.Moreover, the time responses of each controlled axes werevery quick. In addition, according to the result of levitatedrotation characteristics, stable levitated rotation was observed



in each actively controlled axes up to 2000 min−1.Future works will be continued to obtain higher perfor-

mance. The experimental setup will be also be modified forspecific applications such as artificial heart and/or ultrapurewater pump.

References

( 1 ) Y. Okada and K. Nonami: “Research Trends on Magnetic Bearings”, JSMEInt. Journal, Series C, Vol.46, No.2, pp.341–342 (2003)

( 2 ) A. Chiba, T. Fukao, et al.: “Magnetic Bearings and Bearingless Drives”,Newnes (2005)

( 3 ) Y. Okada, T. Ohishi, and K. Dejima: “General Solution of Levitation Con-trol of a Permanent Magnet(PM)-Type Rotating Motor”, JSME Int. JournalSeries C, Vol.38, No.3 (1995)

( 4 ) R. Schoeb and N. Barletta: “Principle and Application of a Bearingless SliceMotor”, JSME Int. Journal Series C, Vol.40, No.4 (1997)

( 5 ) Y. Okada, S. Ueno, T. Ohishi, T. Yamane, and C.C. Tan: “MagneticallyLevitated Motor for Rotary Blood Pumps”, Artificial Organs, Vol.21, No.7,pp.739–745 (1997)

( 6 ) S. Ueno and Y. Okada: “Characteristics and control of a bidirectional axialgap combined motor-bearing”, IEEE/ASME Transactions on Mechatronics,Vol.5, No.3, pp.310–318 (2000)

( 7 ) K. Shimbo, I. Tomita, O. Ichikawa, C. Michioka, A. Chiba, and T. Fukao:“Axial Gap Length and the Maximum Torque of Shaftless Axial Gap Bear-ingless Motors”, IEEJ Annual Meeting 1997, Paper No.1218 (1997) (inJapanes)

( 8 ) K. Chongkwanyuen, O. Ichikawa, C. Michioka, A. Chiba, and T. Fukao: “In-clination Control of Axial Gap Bearingless Motors”, IEEJ Annual meeting1997, Paper No.1219 (1997) (in Japanese)

( 9 ) M. Osa, T. Masuzawa, and E. Tatsumi: “Miniaturized Axial Gap MaglevMotor with Vector Control for Pediatric Artificial Heart”, Journal of JSAEM,Vol.20, No.2, pp.397–403 (2012)

(10) H. Takada, N. Kurita, and T. Ishikawa: “Proposal of a Double tator TypeMagnetically Levitated Axial Gap Motor”, IEEJ Industry Applications Soci-ety Conference 2012, Paper No.Y-114 (2012) (in Japanese)

Nobuyuki Kurita (Member) received B.S. degree and M.S. degreefrom Ibaraki University in 2001 and 2003, respec-tively. And he received his Ph.D. degree in engi-neering from Ibaraki University in 2006. He joinedGunma University as an assistant professor in 2009.His research interests include application of magneticbearings and self-bearing motor. Dr. Nobuyuki Kuritais a member of IEEE and IEEJ.

Takeo Ishikawa (Senior Member) graduated from Tokyo Institute ofTechnology in 1983. He joined to Gunma Universityin 1983, and now he is a professor. His research in-terest includes electrical machine and power electron-ics. He received the 1998 best paper award of IEEETransaction on Vehicular Technology.

Hiromu Takada (Non-member) received B.S. from Gunma Univer-sity in 2010. He is Master student of graduate schoolof Gunma University. His research interests includeapplication of magnetic bearings and self-bearingmotor.

Genri Suzuki (Non-member) received B.S. from Gunma Universityin 2012. He is Master student of graduate school ofGunma University. His research interests include ap-plication of magnetic bearings and self-bearing mo-tor.


proposal of a permanent magnet hybrid-type axial

Documents